Summary of Abstracting Sparse Dnn Acceleration Via Structured Sparse Tensor Decomposition, by Geonhwa Jeong et al.
Abstracting Sparse DNN Acceleration via Structured Sparse Tensor Decomposition
by Geonhwa Jeong, Po-An Tsai, Abhimanyu R. Bambhaniya, Stephen W. Keckler, Tushar Krishna
First submitted to arxiv on: 12 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Hardware Architecture (cs.AR)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper addresses the challenge of accelerating deep neural networks (DNNs) while exploiting sparsity. It proposes tensor approximation via structured decomposition (TASD), a method that transforms any sparse tensor into a series of structured sparse tensors, allowing for hardware acceleration. The TASDER software framework is developed to accelerate DNNs by searching layer-wise high-quality structured decompositions for both weight and activation tensors. This approach can accelerate off-the-shelf dense and sparse DNNs without fine-tuning, resulting in energy-delay-product improvements of up to 83% and 74%. The proposed method bridges the gap between sparse DNN models and hardware, enabling efficient acceleration on various systems with structured sparse hardware support. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps make computers faster by speeding up special kinds of computer programs called deep neural networks. These programs are like super powerful brains that can do lots of things like recognize pictures or understand speech. The problem is that these programs take a lot of work to run, and it’s hard to make them run fast on most computers. To fix this, the authors came up with a new way to break down these programs into smaller parts that are easier for computers to handle. This makes it possible to use special computer chips called structured sparse hardware that can speed up the program without needing to change the program itself. The authors also created software to help make this work better on different kinds of computers. |
Keywords
* Artificial intelligence * Fine tuning
